An Optimization Model for Greenhouse Growth Conditions Using Machine Learning Techniques
摘要
To meet the growing worldwide food needs, sustainable agricultural practices are necessary, with greenhouse maximization being a leading method. This research examines the use of machine learning algorithms in optimizing greenhouse growth conditions with the Advanced IoT Agriculture 2024 dataset containing chlorophyll, plant height, biomass, root measurements, and more. For adaptive crop management, three models were devised to calculate optimal environmental conditions—Random Forest, XGBoost, and Neural Networks. Environmental parameter estimation with Random Forest gave the most accurate results with an MAE of 0.12 and R2 of 0.92. Followed by XGBoost with MAE = 0.20 and R2 = 0.89. Neural Networks seemed to be overfitting and required better generalization. Through feature importance analysis, average chlorophyll content (ACHP) and average daily water vapor (ADWV) were forecasted to have the highest impact. The findings underscore the potential value of applying AI in precision agriculture; however, model generalizability, changing environmental conditions, and computational limitations still pose significant hurdles.